Seventy spring brood specimens of Leptidea spp. were collected in the vicinity of Novosibirsk Academy Town, Novosibirsk Province. The collection area was a 100–200 m wide and a 4.5 km long continuous stripe of meadows adjacent to birch/ pine forests, along the bank of the Novosibirsk Water Reserve and the parallel railroad, between Obskoe More railway station (54°47'37"N; 83°04'34"E; (DMS)) and a point (54°50'04"N; 83°04'40"E (DMS)), 900 m NNE of Rechkunovka railway station, at elevations of 107–137 m a.s.l. (see the locality on a schematic map of northern Eurasia in Fig. 2). Note that the northern half of this area belongs to the city of Novosibirsk while the southern half to the satellite town of Berdsk (with the border at Beregovaya railway station in the middle of the collecting area). Berdsk is the type locality of the subspecies Leptidea reali yakovlevi Mazel, 2003 (the justification of which to our opinion have been insufficiently reasoned in the original description). The specimens were collected by O.E. Kosterin in June 2010, May 2011 and May 2012 with a net and frozen immediately. Details of the specimens examined are provided in Table 1. In screening for Wolbachia infection, the combined sample of L. juvernica and L. sinapis was updated with 15 more specimens which were not identified to either of these two species, not analysed in other respects and not included into Table 1, so that the total sample of Wolbachia screening contained 85 Leptidea specimens.

Figure 2.

Position (black circle) of the studied area at the border of Novosibirsk city and Berdsk town (54°47'37"N; 83°04'34"E – 54°50'04"N; 83°04'40"E; DMS), Novosibirsk Province, Russia, on a schematic map of northern Eurasia.

Genomic DNA was extracted according to Bogdanova et al. (2009), with modifications for isolation from individual insects. Frozen specimens without abdomen and wings were homogenized in 0.6 ml 0.15 M NaCl. The homogenate was centrifuged (3,300 rcf, 5 min) and the supernatant was discarded, then 0.2 ml solution for DNA extraction (0.1 Tris-HCl, pH 8.0; 5 mM EDTA; 0.5% SDS; 0.1 M NaCl) was added and incubated at room temperature for 40 minutes. Then the solution was centrifuged (16,100 rcf, 5 min) and the pellet was discarded. To remove proteins and RNA, LiCl (0.2 ml, 5M) was added to the supernatant solution and incubated on ice for 15 min. The solution was centrifuged (16,100 rcf, 5 min) and the supernatant was transferred to fresh tubes. Ethanol (1 ml, 96%) was added and the mixture was incubated on ice for an hour. Then it was centrifuged (16,100 rcf, 10 min) and the supernatant was discarded. The precipitate was washed with 0.1 ml 75% ethanol and centrifuged (16,100 rcf, 5 min), then dried at 50 °C for 5 min and dissolved in 50 μl of deionized H₂O.

A 708 bp long fragment of the COI gene, positions 1526–2156 (positions are given according to the mitochondrial reference of Drosophila yakuba Burla, 1954 (AN X03240)), was amplified with the universal insect primers LCO-1490 and HCO-2198 (Folmer et al. 1994). A 684 bp sequence of ITS2 (internal transcribed spacer 2) and 571 bp sequence of CAD (carbamoyl phosphate synthase II, Aspartate carbamoyltransferase, dihydroorotase) were amplified with primer pairs ITS3/ITS4 and CADFa/CADRa, respectively, following Dincă et al. (2011). The H1 gene was amplified with primers designed for two overlapping sequences: the 5’ terminal part with the primer pair LH4-f (5’ACCCTGTACGGTTTCGGCGGTTAA) and HeH1-r (5’ AGCGCCCTTGCCCTTGGTCTGTATC) and the 3’terminal part of gene with another pair, HeH1-f (5’ACCCACCCCAAGACCTCCGAGATGGT) and LeH1C-r (5’AGGGGGACTCACTTTTTGGA). The 5’ terminal fragment is approximately 1.5 kbp long, and the 3’-terminal fragment is 650 bp long. The primers were originally designed to match orthologous sequences of Bombyx mori (Linnaeus, 1758) (LH4-f), Heliconius erato (Linnaeus, 1758) (HeH1-r and HeH1-f) (Solovyev et al. 2015) and L. sinapis (LeH1C-r); they were produced by Biosset (Novosibirsk. Russia).

DNA samples were examined for Wolbachia infection by amplification of wsp with the following primer set: wsp81F (5’TGGTCCAATAAGTGATGAAGAAAC-3’), wsp691R (5’AAAAATTAAACGCTACTCCA-3’) (Braig et al. 1998). PCR products of five DNA stocks of four species were sequenced.

PCR mixtures (30 μl) contained 0.2 mM of each dNTP, 1.5 mM MgCl₂, 25mM KCl, 60 mM Tris-HCl (pH 8.5), 10 mM β-mercaptoethanol, 0.1% Triton X-100, 0.5 µM of each primer, 1 μl of genomic DNA solution and 1 U of Taq DNA polymerase or 1 U of Smart-Taq DNA Polymerase (by Laboratory Medigen, Novosibirsk, Russia). PCR was performed using a thermal cycler MyCycler (Bio-Rad, USA) with the following program: 1) 94 °C — 2 min 30 s, 1 cycle; 2) 95 °C — 15 s, 47–55 °C — 30 s, 68 °C — 1 min, 35 cycles; 3) 68 °C — 2 min, 1 cycle.

The entire coding sequence of H1 and a 631 bp long fragment of COI (positions 1526–2156) were sequenced. The Sanger reaction was conducted in 30 μl volume of mixture containing 1 μl of BigDye Terminator, version 3.1 (Applied Biosystems), 100–200 ng of DNA, 3 pmol of primer and 6 μl of buffer solution for BigDye 3.1. A MyCycler (Biorad) thermocycler was used with the following program: 95 °C — 45 s, 50 °C — 30 s, 60 °C — 4 min; 26 cycles. Sequencing was made at the SB RAS Genomic Core Facility, Novosibirsk.

Sequence alignments and calculation of the genetic distances were performed using the MEGA 5.0 software package (Tamura et al. 2011).

For genotyping the L. sinapis and L. juvernica specimens with respect to certain diagnostic nucleotide substitutions in mitochondrial and nuclear markers, CAPS analysis was conducted (Konieczny and Ausubel 1993). After the analysis of DNA sequences of L. sinapis and L. juvernica in public databases, we picked the set of endonucleases HpaII, AluI and HindIII for genotyping the COI gene, ITS2 region and CAD gene, respectively.

The 708 bp long amplified fragment of COI of L. juvernica contains three restriction sites for endonuclease HpaII and is digested to 4 fragments (66, 109, 206, 327 bp), while the orthologous fragment of L. sinapis has no restriction sites. The ITS2 region of L. sinapis contains the only site specific for endonuclease AluI and is digested into 2 fragments (412, 272 bp); the ITS2 of L. juvernica does not contain restriction sites for AluI. The CAD sequence of L. juvernica includes only one restriction site for endonuclease HindIII, which digests it into two fragments, 110 and 461 bp in length; the CAD sequence of L. sinapis has two sites which produce three digestion fragments (110, 189, 272 bp). The buffers and enzymes for restriction reactions were produced by Sibenzim, Novosibirsk, Russia. The identical procedure was used for different markers, as follows: 9 μl of the PCR product was added with 0.5 U of endonuclease and 1 μl of a buffer relevant to the endonuclease. The mixture was incubated at 37 °C for 2 hr, inactivated at 80 °C for 20 min and analyzed by electrophoresis in 1.5% agarose.

The abdomen tip with the genitalia was taken from frozen specimens of L. juvernica and L. sinapis, incubated for 10 min at 98 °C in 10% potassium hydroxide for maceration and dissected under a stereomicroscope. Lengths of the valve (V) and saccus (S) were measured with an ocular-micrometer and binocular lens MBS-2, as shown in Fig. 3. Besides, the saccus curvature was classified as referring to arbitrary binary scores: 0 – straight, 1 – S-like curved.

Figure 3.

Male (a specimen L12) and female (b specimen L45) genitalia of Leptidea juvernica, with measured parameters indicated, namely the length of the saccus (S), valve (V) and ductus bursae (D).

Statistical analyses were carried out using MS Excel 10 for Windows.

The genitalia were analysed before molecular analysis, which was carried out blindly of the genitalic results. The specimens in which molecular results appeared discordant with morphological ones, were then rechecked for morphology and discordancy was confirmed.

The external characters reported to be different in the spring brood males of L. juvernica and L. sinapis, namely (1) the wing underside below vein M3 more suffused by dark scales and with less expressed lighter spots between veins and (ii) more attenuate fore wing apex in the former species (Ivonin et al. 2009), are difficult to measure and somewhat subjective. Therefore we classified them as referring to arbitrary binary classes:

• hind wing underside suffusion below vein M3: 0 – with well-expressed lighter spots between veins, 1 – stronger, more even, with scarcely or expressed lighter spots;

• fore wing apex: 0 – broadly rounded; 1 – more attenuated and acute (Fig. 4).

Figure 4.

Males of Leptidea sinapis (a specimen L1 b specimen L26) and Leptidea juvernica, (c specimen L3 d specimen L8), with different scores of subjectively evaluated wing characters: the hind wing underside suffusion below vein M3: 0 – lighter, with better expressed lighter spots between veins; 1 – stronger, with scarcely seen lighter spots; and the fore wing apex shape: 1 – more acute, 0 – more rounded. The scores for the shown specimens are as follows (suffusion, apex shape): a (0,0); b (1,0); c (1,1); d (1,0); a and c are variants most frequent in the respective species.

The 631 bp long fragment of the mitochondrial gene COI (position 1526 – 2156) was sequenced for 34 Leptidea specimens (10 of L. sinapis + L. juvernica, 11 of L. morsei and 13 of L. amurensis) collected in the same locality. The sequences were submitted to European Nucleotide Archive (ENA), for accession numbers see Table 1. The sequences of L. morsei specimens were identical. In L. amurensis, two alleles were found which differed in position 1969, occupied by either T or A. The 1969T allele was found in 12 specimens while the 1969A allele was only found in only specimen L57. Six COI alleles were revealed in L. sinapis and L. juvernica. These six alleles differed in 22 sites (Table 2) and formed two groups of three alleles each, further referred to as the s- and j-alleles. The consensuses of each group differed in 17 substitutions, the other 5 substitutions were not diagnostic. The j-alleles differed from each other in substitutions in the positions 1615, 1686, 1917. Two of the s-alleles differed in T/A substitution in position 2076, while the third, found in specimen L15, has positions 1587 and 1674 occupied by the nucleotides otherwise specific for j-alleles. Hence the two latter positions were not diagnostic. As a result, the set of positions diagnostic for the s- and j-type, which allows species identification, included 15 positions (Table 2).

The averaged and minimum p-distances between of the studied COI fragment between L. juvernica (j-alleles), L. sinapis (s-alleles) L. morsei, L. amurensis, of are provided in Table 3.

PCR amplification of the wsp gene revealed Wolbachia infection in 38 of 42 tested males and 18 of 19 tested females of the L. sinapis + L. juvernica united sample (91.8% prevalence), in all 11 tested specimens of L. morsei and in all 13 tested specimens of L. amurensis (100% prevalence) (Table 1, but 15 specimens of L. sinapis or L. juvernica, analysed only for wsp, are not included into the table).

The wsp gene was sequenced in one specimen of each L. amurensis (L58), L. morsei (L67) and L. juvernica (L12) and two specimens of L. sinapis (L16, L17). The sequences were submitted to the PubMLST database http://pubmlst.org [accessed 30 January 2015] (for accession numbers see Table 1). L. amurensis, L. juvernica and one specimen (L17) of L. sinapis turned out to have allele wsp-10, with the following hypervariable regions: HVR1-10, HVR2-8, HVR3-10, HVR4-8. The specimen L16 of L. sinapis had wsp-687 allele, which differed from wsp-10 with one non-synonymous nucleotide substitution A193G (serine to glycine). This allele had not been previously recorded and was designated at http://pubmlst.org as wsp-687. L. morsei had a Wolbachia strain with another new allele, designated as wsp-688. This allele differed from wsp-10 with 81 nucleotide substitutions (uncorrected p-distance 0.169) and gaps, resulting in 41 amino acid differences, and had the following hypervariable regions: HVR1-2, HVR2-267, HVR3-2, HVR4-23.

The CAPS approach (see ‘Materials and methods’) allowed us to test 36 more specimens of L. sinapis/L. juvernica in addition to those 10 in which COI was sequenced. We distinguished s- versus j-alleles of the mitochondrial marker COI and nuclear markers CAD and ITS2 in the same set of specimens. The three sets of CAPS data, for all three markers, were fully concordant: each specimen possessed either only s- or only j-alleles for all three markers. This gave us a reason to consider and further refer these specimens as belonging to the true biological species L. sinapis and L. juvernica, respectively.

The 747 bp long coding sequence of the H1 gene of histone H1 was sequenced in 9 specimens: L10, L12 (L. juvernica), L15–L17 (L. sinapis), L50, L58 (L. amurensis), L60, L67 (L. morsei); the sequences were submitted to ENA (for the accession numbers see Table 1). L. sinapis and L. juvernica appeared to have identical primary structure of the H1 coding sequence. Comparison of those of L. sinapis, L. juvernica, L. morsei and L. amurensis revealed 10 polymorphic sites, seven of which reside in the region coding for the C-terminal domain. As compared to the consensus H1 coding sequence for all the four species, the H1 sequence of L. morsei has two transitions, G570A and A654G, while that of both L. sinapis and L. juvernica has two transitions, G27A and G63A, and two transversions, C456G and A648C. H1 of L. amurensis has three transitions G36A, G346A, G456A, and 1 transversion C453G. The substitution G346A was in the codon first position and lead to the amino acid substitution A116T, while all other above mentioned substitutions are in the third positions and synonymous. Besides, the sequenograms of both studied specimens of L. amurensis showed in position 306 overlapping peaks for G (as in the consensus) and T (synonymous substitutions). In one of those specimens (L50), analogous simultaneous presence of both C and T was revealed in position 219. This could result from either heterozygosity for two alleles in homologous histone gene clusters and/or cis-heterogeneity for the repeated H1 copies in the same cluster.

The lengths of the following genital structures were measured: the saccus and valve in males (Tables 4 and 5) and the ductus bursae in females (Tables 4 and 6). Besides, in males, we qualitatively evaluated additional characters such as the shape of saccus (straight versus S-like curved) and some wing characters (Table 7). Females of these species did not differ in external characters.

Two classes of spring brood females of the s- and j-groups with respect to the ductus bursae length were concordant with the CAPS data. The mean ductus length was significantly (p<.001) inferior in the s-group, and the length distributions of these groups did not overlap (Tables 4 and 6). In males, the difference between groups in the mean lengths of the saccus and valve were significant as well, with p<.001 and p<.01, respectively (Table 4). At the same time, the distributions of both the saccus and valve lengths of the s- and j-groups overlapped (Fig. 5).

Figure 5.

The saccus and valve lengths of the L. sinapis and L. juvernica. A Plot of the saccus length against the valve length of L. sinapis and L. juvernica, as identified by molecular markers B Plot of the ratio of the saccus length to the valve length against the saccus length for the same sample.

The saccus curvature did not appear as a reliable differentiating character as well, since its mean square contingency coefficient (φ coefficient) value with the CAPS data was rather small (φ = 0.50, p<.001).

The size, coloration of the hind wing underside and the shape of apex of the fore wing were also found to associate with the molecular groups j and s but again with small values of the φ coefficient: φ = 0.49 for the size, φ = 0.73 for the hind wing coloration p<0.001; φ = 0.29 for the fore wing apex p<0.05.

It may be concluded that neither the genital structure lengths, nor the saccus curvature, nor the general size, nor the wing coloration allow reliable identification of males of the s- and j-groups.

The observed differences in the studied COI fragment of L. sinapis and L. juvernica are substantial. They are illustrated by the averaged and minimum p-distances provided in Table 3 (both values being very close to each other). This result well agrees with the earlier published data (Dincă et al. 2011, 2013, Lukhtanov et al. 2011). Full concordance of alleles of two unlinked nuclear genes and mitochondrial genes unequivocally supports the existence of integrated “molecular species” which can be identified based on diagnostic nucleotide substitutions in either of the mentioned genes, by CAPS-analysis or direct sequencing. Since any interspecies cross would bring about discordance of these markers, that we did not detect, the gene flow between these ‘molecular species’ is either absent or very limited. Thus, our ‘molecular species’ are at the same time valid biological species according to the Mayerian species concept. They are to be identified as taxonomical species Leptidea sinapis and L. juvernica, according to the predominating morphological characters used to be considered diagnostic for species bearing these names. At the same time, we claim that these morphological characters are not satisfactory for species identification, since opposite variants of each of them are still present in each of the two species.

In some studies the task of quick and still reliable identification of species of the L. sinapis complex by application of a morphometric approach was achieved with a 100% efficiency (Fumi 2008, Sachanowicz 2013). In other cases, overlapping of morphometric characters was observed so that some specimens could not be unequivocally identified (Hauser 1997, Kudrna 2001, Verovnik and Glogovčan 2007). This could result from either insufficient genetic isolation of species or a greater intraspecific variation, e.g. driven by ecological factors (Bolshakov et al. 2013, Tsvetkov 2007, Fumi 2008) or from differences between the spring and summer brood (Sachanowicz 2013). Fumi (2008) also noted the potential effect of choosing poor diagnostic characters and measurement errors.

According to the discriminant criterion suggested by Fumi (2008) for females of L. sinapis and L. reali, the critical value for the ductus bursae length was 0.79 mm. According to our data, the hiatus of this character between females of L. sinapis and L. juvernica is at the interval of 0.60–0.75 mm.

The length of the saccus and valve in our case appeared insufficient for a complete discrimination of L. sinapis and L. juvernica. The ratio of these values, which allowed 98% discrimination of males of L. sinapis and L. reali (Fumi 2008), in our case also had better resolution but did not yet display a hiatus between the species (Fig. 5). Ideally, the way of discriminating should be unequivocal and not depend on geographical, ecological or seasonal circumstances. Fumi (2008) achieved 100% discrimination of males of L. sinapis and L. reali through simultaneous analysis of four genitalic characters: the lengths of the aedeagus, saccus, valve and uncus, while adding further characters to the multivariate analysis did not further contribute to resolution. However, discriminative analysis of several morphometric characters of the genitalia is laborious and hence impractical for routine identification of specimens. At present, molecular analysis involving either of our CAPS-markers is an easier means of identification a specimen than multivariate analysis of the genitalia morphology. Furthermore, the differentiation of L. reali and L. juvernica is still beyond the morphological approach and until now these species can be distinguished only by molecular markers and/or karyotype (Dincă et al. 2011, 2013).

Other external characters, such as the general size and coloration of the hind wing underside, recognised by a naked eye, are unreliable and allow only a first approach to species identification in the field (Ivonin et al. 2009), as follows from the low values of the φ coefficient for association of these characters with molecular markers. This may result from the conventional nature of the character grades or from the great variability for these characters, maybe as a remnant of introgression between species in the past. Note, however, that no males of L. juvernica (identified by molecular markers) were scored as ‘0’ as to the hind wing underside suffusion (that is with well-expressed lighter spots below vein M3), the variant found in the majority of males of L. sinapis (Table 7). This somewhat agrees with the data by Ivonin et al (2009) who did not find external characters of L. sinapis among the males of L. juvernica identified by the genitalia.

Anyway, we conclude that molecular and karyological characters are so far the only reliable means of identification of L. sinapis and L. juvernica, and the molecular ones are much easier methodically.

Divergence and fixation of alleles of genes responsible for reproductive isolation are sufficient for speciation (Wu 2001). If genitalic differences contribute to reproductive isolation, genes governing the genital structures are to occur among them quite often. Differences in the genitalia size and structure result from the realization of the ontogenetic program. Since Leptidea sinapis and L. juvernica are closely related species, they should have the same set of orthologous genes with a similar system of expression regulation in ontogenesis. A new allele in one of such ‘genital’ genes which once occurred in a small population can soon be fixed by gene drift, giving rise to a nascent genetically isolated species. Developmental genes often have pleiotropic effects and their mutation can bring about changes in a complex of morphological characters. In particular, the same gene may affect the length of both male and female genitalia through an effect on the size of the anlagen of genital organs in early embryogenesis in both sexes. Hence, the differences in the male and female genitalia between L. sinapis and L. juvernica may be determined by the same single major gene but also be influenced by small effects of an unknown number of other genetic and/or environmental factors. This seems to be common in Leptidea: thus, Leptidea lactea Lorkovic, 1950 was isolated from L. morsei because of an observed bimodal distribution of the genitalia length in united samples (Lorcovic 1950).

The overlap of distribution of the length of the male genital structures may be interpreted through presence in both species of both ‘long’ and ‘short’ alleles of the hypothetical gene responsible for differences between L. sinapis and L. juvernica, although with oppositely biased frequencies. This could result from:

• inheritance of both alleles from the common ancestor,

• introgression between species, and

• de novo mutational re-appearance of ‘long’ and/or ‘short’ alleles.

Introgression is a common phenomenon for sympatric closely related butterfly species. According to an estimation by Mallet (2005), about 16% of 440 European butterfly species can hybridise with at least one other species in natural conditions. Most of such hybrids, especially females, suffer from lowered fertility or complete sterility in F1. However, some interspecific hybrids are still able of backcrossing with one of the parental species, which can lead to gene flow in hybrid zones (Mavárez et al. 2006, Descimon and Mallet 2009).

Specimens from Novosibirsk Province with intermediate state of diagnostic external characters or, more frequently, with discordant combination of characters of L. sinapis and L. juvernica were supposed to be interspecies hybrids or products of their backcrosses (Kosterin et al. 2007, Ivonin et al. 2009). In particular, Ivonin et al. (2009) claimed that among the spring brood males with the genitalia of L. sinapis, there were specimens with the external characters of L. juvernica (a dark, evenly suffused hind wing underside below vein M3, a processed fore wing apex). Oppositely, the males with L. juvernica genitalia were homogenous for these characters. Our data do not support the latter claim, for the coloration of hind wing underside and the fore wing apex shape varied strongly in males of both L. sinapis and L. juvernica.

An attempt to reveal hybridization between L. sinapis and L. reali in the Pyrenees using 16 allozyme loci was unsuccessful (Martin et al. 2003). Verovnik and Glogovčan (2007) suspected a possible hybridisation between L. sinapis and L. juvernica in Slovenia. They revealed some unusual specimens which either had the saccus of an intermediate length or had a long saccus but short aedeagus. These specimens referred to L. sinapis according to molecular markers. The same authors revealed two specimens morphologically corresponding to L. juvernica but belonging to L. sinapis according to the COI sequence. However, their RAPD analysis revealed fragments specific both to L. sinapis and L. juvernica, thus suggesting the hybrid nature of those specimens. The most recent large scale study did not detect signs of introgression among the three species of the L. sinapis complex (Dincă et al. 2013). On the contrary, the existence of biochemical and behavioural prezygotic barriers among them were demonstrated (Friberg et al. 2008b, Dincă et al. 2013).

Inheritance of ‘long’ and ‘short’ alleles in the common ancestor of L. sinapis and L. juvernica is also a plausible interpretation. These alleles could be involved in genetic isolation of the nascent species by forming a reproductive barrier between them, but fixation of either allele in these species may not have taken place. The initial genital prezygotic barrier could later be strengthened by adding biochemical and behavioral barriers. They would lower significance of the primary genital barrier and somewhat release selection for the ‘long’ versus ‘short’ alleles and vice versa, allowing their frequency to drift.

At this stage of our knowledge, the third scenario of arising ‘long’ and/or ‘short’ allele(s) cannot be excluded as well.

In contrast to core histones, histone H1 is a very variable protein (Berdnikov et al. 1993, Happel and Doenecke 2009). For this reason its gene served well for reconstructing phylogeny of the genus Pisum L. (peas) both at inter- and intraspecies level (Zaytseva et al. 2012, 2015). In spite of its variability elsewhere, the H1 gene appeared identical in L. sinapis and L. juvernica. The H1 gene variation revealed in the four studied species of Leptidea turned out to be five times lower than that of COI, differing from our data obtained for three species of Oreta Walker, 1955 (Drepanidae) (Solovyev et al. 2015), where the substitution rate in H1 appeared to be only twice less than that in COI.

All the four studied Leptidea species were found to be infected with Wolbachia, with prevalence of infected specimens of 91.8% in L. sinapis + L. juvernica and 100% in L. amurensis and L. morsei (Table 1). We cannot exclude the possibility that all individuals in the studied populations are infected. We removed the abdomen and hence isolated DNA from somatic tissues only, while the Wolbachia presence can be limited to reproductive tissues (Dobson et al. 1999). The high level of infection probably indicates at functional effect of Wolbachia on the host, ranging from mutualism (increase of the host fitness) to a reproductive parasitism (cytoplasmic incompatibility, feminization, parthenogenesis, male-killing) (Werren 1997; Werren et al. 2008). We can exclude feminisation, male-killing or parthenogenesis that would result in biased sex ratio, which was not observed (Table 1). Special experimental studies would be necessary to investigate the reason of the high infection level in these four Leptidea species.

Wolbachia infection is vertically transmitted through host generations via maternal cytoplasm. Therefore the phylogeny of Wolbachia could be expected to be concordant with the phylogeny of its hosts. However, Wolbachia can as well be transmitted horizontally between related species through introgression and between unrelated species by unknown agents. In addition, Wolbachia strains as well as their particular genes such as wsp may result from recombination between different strains.

We found three Wolbachia strains in Leptidea according to the wsp gene sequences. Three species, L. amurensis, L. sinapis and L. juvernica, were found to possess allele wsp- 10 which is widespread in insects. According to our counts at the http://pubmlst.org, it was so far registered in 27 species of Lepidoptera from different families (Pyralidae, Hesperiidae, Papilionidae, Pieridae, Nymphalidae, Lycaenidae) and, and also in Culex pipiens Linnaeus, 1758 from Diptera. The second allele wsp-687 was found in L. sinapis for the first time. It differs from wsp-10 in only one nucleotide substitution. The third wsp-688 allele was also found for the first time, in L. morsei. This new allele has a unique hypervariable region 2, HVR2-267, while other hypervariable regions HVR1-2, HVR3-2, HVR-23 were found elsewhere (Baldo et al. 2005, 2010; Baldo and Werrein 2007). Nevertheless the mentioned combination of the hypervariable regions has not been so far recorded, thus the allele is a product of recombination of strains representing different evolutionary lines of Wolbachia.

The pattern of Wolbachia variants in the studied Leptidea species is discordant to the host phylogeny. Variation of wsp sequences in L. sinapis, L. juvernica and L. amurensis is extremely low, viz. wsp-10 allele is common for these species and a closely related wsp-687 allele is also found in L. sinapis, whereas L. morsei possesses a highly divergent wsp-688 allele.

The wsp-10 allele could hardly be inherited from the common ancestor of the three species, taking into account a considerable degree of variation accumulated by the host Leptidea genes, both nuclear and mitochondrial (Dincă et al. 2011, 2013 and this paper). Note that supposition (i) contradicts also the phylogenetic relationships of the species involved (Dincă et al. 2011) as follows: the branch L. morsei + L. amurensis is opposed to the branch containing L. sinapis and L. juvernica. The low wsp variation in three species may have two explanations:

1. the same strain of Wolbachia could have spread across the three species by interspecies crosses;

2. The same wsp allele could have spread across the three species via horizontal transfer of Wolbachia.

We exclude option (i) since we failed to trace such crosses by other molecular means. Explanation (ii), that is independent infection by the same Wolbachia strain, is most probable because of a high frequency of wsp-10 in butterflies. L. morsei was no doubt independently infected by an unusual Wolbachia strain with wsp-688, however, more data on L. morsei is necessary to consider the evolutionary history of its Wolbachia.